Event-year risk-indicator dataset and dual-channel case-retrieval code for 13 transnational renewable energy projects

Transnational renewable energy projects under information scarcity face cascading geopolitical, economic, socio-cultural, environmental, and market risks, where under-preparation costs exceed over-preparation costs by orders of magnitude. Existing decision-support approaches neither reconcile structured indicators with unstructured textual intelligence under high evidential conflict, nor allow case-based reasoning to internalise asymmetric misclassification costs. This study develops an intelligent emergency decision-support system organised as a Sense–Diagnose–Respond–Learn (SDRL) cycle, whose core Diagnose engine comprises (i) a conflict-aware Dempster–Shafer protocol fusing the structured and textual evidence channels, with Murphy correction triggered whenever the conflict coefficient exceeds K = 0.5; (ii) a game-theoretic combination of entropy and analytic-hierarchy-process (AHP) weights; and (iii) a cost-sensitive case-based retrieval algorithm embedding an asymmetric misclassification-cost matrix within the similarity metric, with proven type-preservation, monotonicity, and bounded-influence properties. The system is validated on 13 transnational renewable energy projects whose indicators are collected at each case’s event year from documented public sources. Leave-one-out cross-validation attains a 53.8% strict Top-1 risk-type match rate against a 14.1% random baseline, with the conflict-aware mechanism triggered in all 13 folds (K ≈ 0.87). The recommendation is invariant across the tested ranges of the weighting coefficient and text dimensionality, and ±20% cost-matrix perturbations leave all results unchanged over 200 draws. A diagnostic-uncertainty stress test shows that dual-channel fusion degrades gracefully under risk-type mis specification (53.8% → 46.2%), whereas the cost channel alone collapses (92.3% → 30.8%). The system executes diagnosis and retrieval in seconds and outputs a personalised response plan from a 32-measure standardised library.